concept and laboratory analysis of trajectory-based ...€¦ · analysis of trajectory-based...

Concept and LaboratoryAnalysis ofTrajectory-BasedAutomation forSeparation AssuranceDavid McNally and Chester Gong

An operating concept and a laboratory analysis methodology were devel-oped and tested to examine how four-dimensional trajectory analysis meth-ods could support higher levels of automation for separation assurance inthe National Airspace System. Real-time simulations were conducted inwhich a human controller generated conflict resolution trajectories usingan automated trial plan trajectory generation and analysis function, butonly in response to conflicts detected and displayed by an automatic con-flict detection function. Objective metrics were developed to compare air-craft separation characteristics and flying time efficiency under auto-mated operations with that of today’s operations using common airspaceand common traffic scenarios. Simulations were based on recorded air traf-fic data from the Fort Worth Air Route Traffic Control Center and wereconducted using today’s and nearly two-times today’s traffic levels. Theresults suggest that a single controller using trajectory-based automationand data link communication of control clearances to aircraft could man-age substantially more traffic than under today’s conditions, and with im-proved route efficiency while maintaining separation. The simulation andanalysis capability provides a basis for further analysis of semi-automated,or fully automated, separation assurance concepts.

INTRODUCTION

In its Next Generation Air Transportation System (NGATS) reportthe multi-agency Joint Planning and Development Office (JPDO) de-

The authors are with NASA Ames Research Center, Moffett Field, CA.

Received December 1, 2006; accepted March 13, 2007.

Air Traffic Control Quarterly, Vol. 15(1) pp 35–63 (2007)CCC 1064-3818/95/030163-20

35

scribes an expected two- to threefold increase in air traffic demand bythe year 2025 and the need for new automation technologies andoperating procedures for the National Airspace System (NAS)[JPDO, 2004]. Under today’s operations air traffic controller work-load, severe weather, capacity-constrained airports and other factorslimit airspace capacity and efficiency. In the absence of severeweather, controller workload is probably the single most importantfactor limiting airspace capacity. An air traffic controller’s primarytask is to ensure safe separation by visual and cognitive analysis ofa traffic display and voice communication of control clearances topilots. The Decision Support Tools (DSTs) developed and deployed inrecent years provide trajectory-based information and automationthat assists controllers with conflict detection and resolution andwith arrival flow management. However, the controller still holdsprimary responsibility for safe separation. Though DSTs are provid-ing measurable improvement in today’s NAS [Federal Aviation Ad-ministration, 2003], DST-based concepts and technologies alone arenot expected to support a two- to threefold increase in airspace ca-pacity. Under the NGATS vision, the use of four-dimensional (4D)aircraft trajectory analysis with higher levels of automation for sepa-ration assurance and data link communication are expected to becore components of future airspace operations.

Recently, airspace operating concepts have been proposed for in-creasing airspace capacity through higher levels of automation and/or delegation of some separation assurance responsibility to the cock-pit. The Advanced Airspace Concept proposes highly automatedseparation assurance for equipped aircraft using air/ground data linkcommunication and an independent safety assurance function [Erz-berger, 2004]. Concepts for delegation of separation assurance to thecockpit through airborne automation and Cockpit Display of TrafficInformation (CDTI) have been proposed [Wing et al., 2004; Barmoreet al., 2004; Battiste et al., 2000], as have concepts that employ a mixof controller-managed and airborne separation assurance methods[Prevot et al., 2005]. The objective of this paper is to make an initialdetermination as to whether or not existing 4D trajectory analysismethods and trajectory-based conflict detection and resolution func-tions have promise as a point of departure for development of thenext-generation separation assurance automation for the NAS.

The concept and laboratory analysis described herein centers onthe following questions: If automated 4D trajectory-based strategicconflict detection and resolution functions could be trusted, could acontroller use them as their primary means to maintain safe sepa-ration? If so, are there resulting benefits in terms of capacity andflying-time efficiency that could be exploited to increase the amountof traffic managed by the controller? And, should the primary stra-tegic conflict detection function fail, could a backup tactical conflict

MCNALLY AND GONG36

detection function [Erzberger, 2004; Paielli and Erzberger, 2005] de-tect an imminent conflict in time to prevent a loss of separation? Forthe purposes of this experiment the operating concept and simulationmethodology assume that aircraft are deviated from their nominalroute of flight or vertical profile only when a traffic conflict is de-tected by the automation. We refer to this concept as “Control byException.” Experimental studies have been previously conducted toinvestigate the controller cognitive workload associated with Control(or “Management”) by Exception operations in air traffic control[Dekker and Woods, 1999]. In the simulations described herein, fol-lowing a detected conflict, the controller uses an automated conflictresolution function to generate a flight plan amendment that resultsin a conflict-free resolution trajectory. Route or altitude restrictions,inter-sector coordination requirements, or sector boundary consider-ations common in today’s operations are not considered when gener-ating resolution trajectories or any trajectory changes. Flight planamendments are transmitted to the aircraft via simulated data linkcommunication.

The paper begins with a description of the simulation methodologythat makes use of existing 4D trajectory analysis methods and Fed-eral Aviation Administration (FAA) en route Center radar track andflight plan data as a basis for higher levels of automation for sepa-ration assurance. The use of actual Center data exposes automationalgorithms and software to a rich variety of real-world traffic condi-tions. The implementation of the Control by Exception concept inhuman-in-the-loop simulation is described, and objective metrics aredefined that provide the basis for a comparative analysis of Controlby Exception operations with today’s operations. The Analysis andResults section describes the various simulation runs that were con-ducted during the study and presents objective comparison of simu-lation operations with today’s operations using common trafficsamples. Simulation runs include conditions where a single control-ler maintains separation in five sectors under nominal and two-timesnominal traffic levels using Fort Worth Center traffic data. The pa-per closes with some concluding remarks. The original, and ex-panded, version of this paper includes conditions where the controllermaintains separation in more than five sectors and includes analysiswith Cleveland Center traffic data [McNally and Gong, 2006].

SIMULATION METHODOLOGY

The simulation methodology is based on an operating philosophywhere all traffic conflicts are assumed to be detected by trajectory-based automation and all trajectory changes, including conflict reso-

TRAJECTORY-BASED AUTOMATION FOR SEPARATION ASSURANCE 37

lution trajectories and pilot-requested route or altitude changes, aregenerated and implemented using trajectory-based automation. Ma-ture trajectory analysis methodologies and software previously de-veloped for DSTs are configured to run such that they automaticallydetect and provide the necessary automation to resolve all trafficconflicts. We refer to this integrated capability for automatic detec-tion and automated resolution as trajectory-based automation (TBA).The human controller relies on the automation to detect and resolveconflicts, but does not scan traffic for conflicts as in today’s opera-tions. Traffic flow and separation characteristics are then measuredand compared to those of today’s operations. This is expected to helpdetermine the suitability of current trajectory analysis methods forhigher levels of automation. It is also expected to help uncover short-comings in trajectory analysis methods that will need to be overcometo achieve the research objectives of two- to three-times traffic den-sity with safety and user-preferred trajectories.

The Center/TRACON Automation System (CTAS) trajectoryanalysis methodology and software are the basis for this analysis[Erzberger et al., 1993; Slattery and Zhao, 1995; Paielli and Erz-berger, 1997; Erzberger et al., 1997; Isaacson and Erzberger, 1997;Erzberger et al., 1999; Swenson et al., 1997]. The CTAS, developed atthe NASA Ames Research Center, includes mature capabilities for4D trajectory prediction, time-based metering, conflict detection, con-flict resolution, flying time analysis of direct routes accounting forwinds, and other functions. The CTAS trajectory analysis and con-flict prediction capabilities are based on real-time analysis of FAA enroute Center radar track (12 s updates) and flight plan amendment(periodic updates) data from the Center Host computer, hourly up-dates of wind predictions from National Oceanic and AtmosphericAdministration (NOAA) Rapid Update Cycle (RUC) model, and adatabase of aircraft performance models. The CTAS conflict detec-tion, trial planning, and direct route advisory functions have beentested extensively under operational conditions at FAA Center facili-ties [McNally et al., 1998, 1999, 2001]. A tactical conflict detectionfunction is being analyzed as a potential enhancement to the legacyHost Conflict Alert function [Paielli and Erzberger, 2005] and is thebasis for the backup tactical conflict alerting function in this analy-sis.

A fundamental requirement of the simulation methodology was ameans for the TBA to provide separation assurance in a simulationderived from actual Center traffic data. To make use of Center trafficdata, a methodology was needed to “undo” the effects of actual con-troller clearances from the traffic data and replace them with trajec-tory changes generated by the simulation controller using the auto-mation to resolve detected conflicts.

The methodology used for this experiment was adapted from a

MCNALLY AND GONG38

previous CTAS simulation experiment [Robinson and Isaacson,2002]. Consider the notional airspace shown in Figure 1, which isdefined by a single airspace sector or any number of adjoining sec-tors. CTAS automation receives a live or recorded feed of actual Hosttrack and flight plan data as described above. Host radar track mes-sages (received every 12 s) are monitored to identify aircraft that areentering the simulation airspace. The radar track update at which agiven aircraft’s Host sector ID (an element of the radar track mes-sage) changes to one of the simulation airspace sectors (i.e., a trans-fer of separation responsibility from a non-simulation sector to asimulation sector) defines the initialization point (IP) for that aircraftin the simulation. At this point, the actual aircraft data source (liveor recorded) is replaced by a simulated aircraft instantiated and con-trolled by the Pseudo Aircraft Simulator [Weske and Danek, 1993].Following the IP, the aircraft simulator generates all subsequentradar track updates based on the initial conditions at the IP (posi-tion, altitude, speed, and flight plan intent) and the NOAA RUC winddata. All subsequent flight plan amendments (route or altitude) aregenerated by the TBA system. Actual Host radar track and flightplan amendments received for the simulated aircraft are ignoredonce the aircraft has passed the IP. Using this methodology, theautomation can be tested using realistic traffic flows derived from anauthentic source of track and flight plan data from any Center forwhich a CTAS adaptation is available.

In this analysis, arrivals to major hub airports in, or near, thesimulated airspace were not converted to simulated aircraft. Insteadthey were allowed to proceed according to their live or recorded trackdata. This simplification was made due to the fact that the conflict

Figure 1. Simulation methodology.


resolution automation was not well configured to solve conflicts be-tween arrivals merging to a common metering fix. Analysis of merg-ing arrival traffic, though an important element of air traffic controloperations near capacity constrained airports, was beyond the scopeof this study. Furthermore, analysis of automation concepts for fu-ture airspace should first be assessed in terms of their ability toaccommodate basic separation assurance problems, i.e., conflict de-tection and resolution problems that do not include arrival metering,before they are extended to problems that require concurrent sepa-ration and metering.

Given that trajectory prediction uncertainties can be on the sameorder of magnitude as safe separation criteria (5 nmi and 1000 ft inen route airspace), uncertainty is an important consideration in theanalysis of TBA concepts. A typical 5–7 m/s (10–14 kn) wind speederror [Schwartz et al., 2000] can result in about a 2 nmi along-trackprediction error over a 10 min time horizon. A similar error in cruisespeed intent can have the same effect. A typical 20% error in theweight estimate for a Boeing 757 aircraft can result in a 3 nmi errorin predicted along-track distance to top of climb over a 10 min timehorizon [Coppenbarger, 1999]. These errors can result in faulty, late,or missed conflict detections and generally require that trajectoryand conflict analysis methods are robust to uncertainty. AggregateCTAS trajectory prediction errors based on analysis of hundreds ofactual en route trajectories can be on the order of 2–3 nmi (1-sigma)for 10 min level flight trajectory predictions, and 1000–2000 ft (1-sigma) for 5 min climbing flight trajectory predictions [Gong andMcNally, 2004]. A number of methods are employed in CTAS to miti-gate the effects of uncertainty including updating trajectory predic-tions at every radar track update, employing expanded separationcriteria for conflict detection (e.g., 7 nmi horizontal and 1500 ft ver-tical for climbing or descending aircraft), and the use of uncertainty-based conflict probability analysis to filter low probability conflicts[Erzberger et al., 1997].

Detailed uncertainty modeling is beyond the scope of this study.However, the methodology in [Gong and McNally, 2004] is applied tocompare the aggregate uncertainty of trajectory predictions basedfully on actual traffic data with those based on simulated traffic datathat is only initialized with actual traffic. The results, presented inthe Appendix, show that the trajectory uncertainty inherent to thesimulation methodology applied in this study is roughly on the sameorder of magnitude as the trajectory uncertainty present under ac-tual traffic conditions. This makes the results more meaningful, sinceuncertainty is an unavoidable aspect of trajectory-based operations.

This study defines the open-loop traffic flow as that which resultswhen all simulated aircraft in the simulation airspace fly an unin-terrupted (i.e., no controller inputs) nominal trajectory along the

MCNALLY AND GONG40

flight plan and vertical profile (e.g., climb profile) that was currentwhen they entered the airspace at the IP. The open-loop simulationprovides a baseline traffic flow in the simulation airspace region towhich metrics can be applied. As shown later, analysis of the open-loop traffic flows provides a quantitative measure of the traffic con-flicts that must be resolved by the separation assurance automationto maintain safe separation in the airspace.

This study defines the closed-loop traffic flow as that which resultswhen the human controller using TBA provides all trajectorychanges needed to maintain separation in the simulation airspace.The TBA includes the CTAS 4D trajectory modeling, conflict predic-tion, trial planning, and graphical user interface functions, and asimple data link communication model. A human controller monitorsa list of predicted traffic conflicts. The controller uses the trial plan-ner automation to manually (and interactively) generate flight planamendments that resolve all traffic conflicts. When the controllerinputs a flight plan amendment, a fixed time delay count down be-gins. Following the time delay count down, the amendment is auto-matically transmitted to the aircraft simulator and to the TBA’sflight plan database. The trajectory database is updated with theflight plan amendment, and the aircraft simulator starts flying theaircraft and generating subsequent radar track updates consistentwith the flight plan amendment.

The fixed time delay simulates the time required for a pilot toreview a flight plan change and indicate compliance or non-compliance. A 24 s time delay seemed plausible for future operationsand was used for all simulation runs in this analysis. In reality thisdelay would vary from a few seconds to a minute or two depending onthe data link. Additionally, a small percentage of flight plan amend-ments would be rejected by the pilot due to aircraft performancelimitations (e.g., too heavy to climb), severe weather avoidance, orother operational factors. A better model could include a random timedelay characterized by a mean and standard deviation and include apercentage of unable responses from pilots requiring that an alter-native resolution trajectory be generated. Requiring a compliancemessage (e.g., will comply, or “wilco”) from the aircraft before theTBA updates the flight plan is expected to be an important aspect ofthe coordination process in trajectory-based operations with datalink communications.

The automatic transmittal of a flight plan amendment to the air-craft simulator combined with the time delay associated with thepilot’s compliance simulates a simple model of data link communica-tion for TBA. This model assumes that consistency in flight planintent (route and altitude) between the aircraft and the TBA systemwill yield adequate consistency between predicted trajectories andactual trajectories. This seemed like a reasonable starting point for


this study. In the future, data link communication could require morecomplete trajectory information to improve consistency between pre-dicted and actual trajectories. Such information could include timeand speed at selected waypoints, speed during certain flight legs(e.g., climb, cruise, descent) and other information. An aircraft’s com-plete preferred fuel-efficient descent trajectory could possibly betransmitted to the TBA system for conflict analysis. Future researchtowards the NGATS vision will determine the information contentand the aircraft conformance requirements for data link communi-cation in trajectory-based operations.

In this analysis, flight plan amendments are generated by a hu-man controller operating the trial planner, but the methodologyshould be suitable for the analysis of any TBA concept.

OPERATING CONCEPT

Figure 2 illustrates the difference between today’s operations and theControl by Exception operations being investigated in this study.Under today’s operations (Figure 2a), the controller monitors a radardisplay showing radar track positions and flight data block informa-tion for all aircraft in the sector. Though DSTs are available to aidconflict detection and resolution, the controller is ultimately respon-sible for detecting traffic conflicts and issuing clearances to maintainsafe separation. Under Control by Exception operations as defined inthis experiment, the TBA detects traffic conflicts and displays themto the controller through a conflict list or other suitable user interfacemechanism as shown in Figure 2b. Because the controller is notasked to identify potential conflicts as in today’s operations, the in-formation traditionally presented in the flight data block is notneeded. Therefore, flight data blocks are not displayed by default.When two aircraft are predicted in conflict by the automation, thecontroller displays their flight data blocks and additional graphicalinformation by a click on the conflict list. Additional pertinentgraphical information is displayed when the controller activates thetrial planner functions [McNally et al., 1998, 1999, 2001] to resolvethe conflict.

As shown in Figure 2b, the controller uses the trial planner tointeractively generate and conflict-probe a trajectory defined by ashallow right turn to an auxiliary waypoint followed by a direct routeto a downstream fix to resolve the conflict. An analysis of the userinterface requirements for Control by Exception was beyond thescope of this study, but the configuration shown in Figure 2b was aworkable nominal display format.

Two operating modes were used during the simulations. Under the

MCNALLY AND GONG42

Figure 2. Controller’s radar display: a) today’s single sector operations, b) five-sector Control by Exception operations.


first, the “Conflict Resolution” mode, a human controller (here, aNASA engineer) used the trial planner to resolve only those conflictsthat were displayed on the conflict list. The controller did not scanand analyze flight data blocks for potential conflicts as in today’soperations. Instead, the controller reacted only to conflicts detectedby the TBA. The controller’s tasks were to: 1) monitor the conflict list,2) use the trial planner to generate route and altitude flight planamendments that would resolve displayed conflicts, and then 3) issueflight plan amendments to aircraft via the simulated data link. Thesetasks were chosen to emulate an automated conflict resolution func-tion, albeit with a human closing the resolution loop. Resolution ma-neuvers were limited to three types: 1) altitude change, 2) directroute to a downstream flight plan fix, and 3) vector to an auxiliarywaypoint followed by a direct route to a downstream flight plan fix(shown in Figure 2b). These are common clearances in today’s airtraffic operations. Speed changes were not used in this experiment.Route or altitude restrictions, or restrictions associated with sectorboundaries or procedural routings, were not considered in the gen-eration of the conflict resolution trajectories.

Expanded separation criteria were used for conflict alerting and forgenerating conflict-free resolution trajectories with the trial planner.Expanded separation criteria provide a safety buffer to guard againstmissed alerts in the presence of trajectory prediction uncertainties.The horizontal separation criterion for conflict alerting was 8 nmi asopposed to the 5 nmi legal horizontal criterion. As an example, theaircraft pair of BTA2574 and N318CT in Figure 2b was listed as aconflict because the minimum horizontal separation was predicted tobe 7 nmi. The vertical separation criterion for alerting was 1500 ft ifat least one aircraft in the conflict pair was climbing or descending atthe point of first loss of separation. If both aircraft were flying levelat the point where horizontal separation was lost, the vertical sepa-ration criterion remained at the legal vertical separation criterion of1000 ft. To provide an additional safety margin when changing tra-jectories, the separation criteria for trial planning were set higherthan that for alerting. The separation criteria for trial planning were12 nmi horizontal for all cases and 2000 ft vertical when one or moreaircraft was climbing or descending.

Under the second operating mode, the “Conflict Resolution & Di-rect-To” mode, the controller resolved all conflicts using the trialplanner as described above and issued all conflict-free direct routeadvisories that were displayed on the Direct-To route advisory list[Erzberger et al., 1999]. The CTAS Direct-To algorithm automati-cally performs a wind-route analysis on all aircraft routings to iden-tify those aircraft that could save at least one min of flying time byflying direct to a downstream fix on their route of flight. Direct-Toroute advisories are limited so as not to propose a route amendment

MCNALLY AND GONG44

that would substantially deviate an aircraft from its nominal route offlight. All direct route advisories are automatically probed for conflictusing the trial plan separation criteria. In the context of this analy-sis, the Direct-To list could be considered to emulate pilot requests acontroller might receive during normal operations. While operatingin this mode, all conflict-free Direct-To route advisories were issuedimmediately without regard for sector boundary or coordination con-siderations common in today’s operations.

METRICS

Minimum Separation Metric

An important objective of this analysis was to develop and applyobjective metrics to compare trajectory-based operations with today’s(baseline) operations. An objective measure of airspace separationcharacteristics was needed to compare the safety and complexity oftraffic under automated operations during the simulations versusthat of today’s operations using a common airspace and a commontraffic sample. In any airspace, the frequency and number of aircraftpairs that pass with a minimum separation that is at, or near, thelegal separation standard (5 nmi or 1000 ft) is one measure of con-troller workload, traffic complexity, and safety. Consider the trafficin any finite airspace, e.g., the five-sector airspace shown in Figure2b. The radar track data for all aircraft that pass through the air-space over a given time interval are analyzed to determine the mini-mum separation distance for each unique pair of aircraft that are notlegally separated by altitude (i.e., those aircraft pairs separated byless than 1000 ft). The minimum separation metric is the numberand frequency of unique aircraft pairs that pass at or near the legalseparation criteria. Plots of this metric for the various simulationruns are shown later in the Analysis and Results section. For anaircraft pair to be considered in the analysis, at least one of theaircraft had to pass through the airspace during the selected timeinterval (this covers cases where one aircraft was in the airspace andthe other was not). For example, over a given interval, say five min,some unique pairs pass with a minimum separation of 10 nmi whileother unique pairs pass with a minimum separation of 50 or 100 nmi.Any unique pair that passes with less than 5 nmi minimum separa-tion while not separated by altitude would reflect a loss of legalseparation and a serious safety violation. The minimum separationmetric is calculated throughout the duration of each traffic scenario.Since this method is based purely on analysis of current-time radartrack data (i.e., predictions are not used), it provides for a simpleobjective comparison of today’s operations with automated opera-tions.


Route Efficiency Metric

The flying time and path distance required for an aircraft to passthrough a given region of airspace is a measure of route efficiency.Measuring route efficiency by path distance alone is not adequatebecause of the effect of wind. Apparent route efficiency gains due toshorter, more direct, routing may not be realized if the TBA does notaccount for potentially unfavorable winds. For this reason, flyingtime is the primary measure of route efficiency. As with the separa-tion metric, a common airspace and traffic sample are used to achievea direct comparison of today’s operations with operations that in-clude higher levels of automation.

Shown in Figure 3 is an aircraft entering a notional simulationairspace region at the initialization point, IP. The Host radar trackhistory reflects the actual aircraft path through the airspace based onHost radar track recordings. The simulated radar tracks reflect thesimulated aircraft’s path through the airspace when the aircraft wasbeing controlled under Control by Exception operations as describedabove. In this example, the actual radar tracks show the aircraftfollowing a standard departure routing though the airspace. Thesimulated tracks show the aircraft flying a shortcut that skips thedeparture routing but, in this case, includes vectoring for traffic.Both actual and simulated track histories in Figure 3 are notionalbut are representative of actual operations. The shortcut and thevector were included in this example to illustrate the point that theefficiency metric uses a common methodology to account for short-cuts, which generally improve efficiency, and vectors for traffic,which generally reduce efficiency.

Figure 3. Route efficiency metric.

MCNALLY AND GONG46

Since the actual aircraft and the simulated aircraft were influ-enced by different wind fields and airspeeds, a method was needed todetermine the equivalent flying time of each aircraft if the winds andairspeeds were the same along their respective routes. Comparingflying times derived by integrating track histories of a given aircraftfrom two different sources is greatly influenced by both airspeed andwind the aircraft is flying through. For example, a simulated aircraftflies through a predicted wind field. The resulting flying time couldvary significantly from the actual aircraft’s flying time, because theactual aircraft was affected by real winds, which likely differed fromthe predicted winds.

Flying time was calculated by integrating groundspeed (true air-speed + wind speed) with respect to path distance along a givenroute. By assuming each aircraft was flying in the same wind fieldwith the same true airspeed profile, any difference in the resultingflying time could then be attributed to the difference in the routing ofeach aircraft. For each track point of a given aircraft, true airspeedwas set to a value determined from the CTAS aircraft model databasebased on aircraft type and phase of flight. The corresponding windspeed was determined from the modeled (NOAA RUC) wind at eachtrack point. Groundspeed is then calculated from the modeled trueairspeed and wind. The flying times resulting from the integration ofthe groundspeed of each aircraft were then compared. Differences areattributed to operational factors affecting routing, e.g., direct routeand/or vectors.

Since irregularities in the geometry of the airspace region couldcause errors in flying-time comparisons, a method was needed toobtain a fair comparison of flying time while considering only thedifferences in operations within the airspace region of interest. Theinitialization point (IP) and the destination airport were the only twopoints assuredly on both actual and simulated trajectories. Routechanges affecting the aircraft while it flew through the simulationairspace were reflected in the track data. The path distance andflying time from the IP to the actual exit point (Exit-A) were com-puted using Host radar track data. The path distance and flying timefrom the IP to the simulated exit point (Exit-S) were computed usingsimulated track data. Once the aircraft exited the simulation air-space, it was assumed the aircraft would continue direct to the des-tination. The specific routes the aircraft took after exiting the simu-lation airspace are irrelevant to this analysis. Therefore, the remain-ing path distance and flying time from the exit point to thedestination airport were calculated for a direct route between the twopoints. The total path distance and flying time, i.e., from IP to Des-tination AP, were the sum of the components inside and outside ofthe simulation airspace. The flight time efficiency metric is the dif-ference between the flying time for the actual route vs. that of the


simulated route. The path distance efficiency metric is the differencebetween the path distance for the actual route vs. that of the simu-lated route.

Flight Plan Amendment Metric

The number of flight plan amendments a controller implementswhile controlling traffic is one objective measure of controller work-load. Using a common airspace and a common traffic sample, thenumber of route and altitude amendments issued while aircraft werein the airspace was compared for simulation operations and actualoperations. Altitude amendments include changes to the plannedflight altitude as well as temporary altitudes. To obtain a fair com-parison, only amendments to aircraft common to both simulated andactual traffic samples were counted. Because descents to satelliteairports within the Center and near-by adjacent Center airports werenot simulated in this analysis, amendments to those aircraft were notcounted.

ANALYSIS AND RESULTS

Conflict Resolution Mode—Today’s Fort WorthCenter Traffic

Five adjoining high altitude sectors in Fort Worth Center (ZFW)airspace (28, 71, 86, 89, and 90) were selected as the simulationairspace. This airspace was chosen because it includes a good mix ofclimbing Eastbound departures from the Dallas/Fort Worth Interna-tional Airport (DFW), climbing North-East-bound departures fromairports in Houston, two arrival streams to DFW, and level over-flight traffic. The simulation and analysis were based on ZFW Hostradar track and flight plan data recorded over a 90-minute periodstarting at 1525 Central Standard Time (CST) on May 26, 2005.

The Host radar track data were first analyzed to establish thebaseline characteristics of the actual recorded traffic flow in the five-sector airspace. Figure 4a shows a time history of the total trafficcount in the five-sector region. Figure 4b shows the minimum sepa-ration metric computed at 5 min intervals over the 90 min recording.Only aircraft pairs not separated by altitude and that passed with aminimum horizontal separation of 10 nmi or less are reflected inFigure 4b. This baseline minimum separation analysis reflects theactions of actual controllers working the traffic under today’s opera-tions. In all of the five-sector airspace, only one unique pair had aminimum separation between 5 and 10 nmi during the 20–25 min

MCNALLY AND GONG48

elapsed time period. Later, during the 60–65 min elapsed time pe-riod, 3 unique pairs had a minimum horizontal separation between 5and 10 nmi. For much of the time, aircraft not separated by altituderemained horizontally separated by more than 10 nmi. Under today’soperations, the number of controllers working this airspace likelyranged from 4 radar (R-Side) controllers when sectors 71 and 90 werecombined (as they often are), to as many as ten controllers when bothR-Side and a D-Side controllers were assigned to each of the 5 sec-tors.

An open-loop simulation of the 90 min, five-sector traffic samplewas conducted to measure the minimum separation metrics thatwould result without any control actions. Recall that the open-loopsimulation methodology, where aircraft fly the flight plan and verti-cal profile as intended upon entering the airspace, effectively re-moves the effects of actual controller actions (e.g., vectors or altitudechanges) from aircraft trajectories in the simulation airspace. Figure5a shows a time history of the open-loop total traffic count. Averagetraffic count varied slightly from run to run (and from baseline toopen-loop), because aircraft exited the simulation region at differenttimes. However, the total number of aircraft entering the simulationregion remained the same for each run. Figure 5b shows the mini-mum separation metric for the open-loop run. Note that Figure 5bshows numerous instances where aircraft violated the vertical and

Figure 4. Separation characteristics, today’s live traffic baseline, five sectors: a)aircraft count, b) minimum separation metric.


horizontal separation standards (dark bars). This was not surprising,given that the traffic flow was effectively uncontrolled. Figure 5billustrates the fundamental reason for the air traffic control system—to maintain safe separation. The number of open-loop loss of separa-tion cases in Figure 5b is directly related to the number of controlactions required to maintain safe separation in the airspace.

A closed-loop simulation was conducted in Conflict Resolutionmode using the 90 min recording in the five-sector ZFW airspace. Asdescribed above, one controller (a NASA engineer) resolved all con-flicts using the conflict list for alerting, the trial planner for resolu-tions, and a simulated data link for communication of control clear-ances to the aircraft. A pilot’s wilco response was simulated by ap-plying a fixed 24 s delay between the time a data link clearance wasissued and the time the simulated aircraft responded.

Sectors 86 and 89 (see Figure 2) include two streams of DFW ar-rivals approaching the metering fix where they enter the DFWTRACON. All DFW arrivals were allowed to run live through thesimulation airspace. In the case of conflicts that were displayed be-tween a DFW arrival aircraft and a non-arrival aircraft, i.e., anyother aircraft in the simulation airspace, the controller resolved theconflict by moving the non-arrival aircraft. This in effect removed thearrival merging and spacing problem from the simulation.

Figure 6 shows a time history of the closed-loop traffic count and

Figure 5. Separation characteristics, open-loop, five sectors: a) aircraft count, b)minimum separation metric.

MCNALLY AND GONG50

the minimum separation metric for the closed-loop five-sector run.Note that the minimum separation metric for closed-loop five-sectoroperations (Figure 6b) is very consistent, arguably a little better,than the minimum separation metric for the baseline run (Figure4b). These data suggest that one controller operating with TBA anddata link communications in a simulation environment can keep anominal flow of aircraft in a five-sector region safely separated.

Actual route and altitude flight plan amendments issued to thedeparture and over-flight aircraft in the five-sector live traffic base-line are shown in Figure 7. It was assumed that flight plan amend-ments issued during the live traffic baseline were to maintain safeseparation or to accommodate pilot requests. On average, approxi-mately seven amendments were issued during each five-min inter-val. There were nearly twice as many altitude amendments as therewere route amendments. Moreover, temporary altitude amendmentsaccounted for 69 of the 84 altitude amendments (82%). Furtheranalysis showed controllers issued multiple temporary altitudeamendments, as many as four, to a given departure aircraft. Occur-rences of multiple temporary altitude amendments may be attrib-uted to the uncertainty associated with conflict detection and reso-lution for climbing aircraft under today’s operations. It is commonunder today’s operations for controllers to issue a temporary altitudeclearance, and input an associated temporary altitude amendment,

Figure 6. Separation characteristics, closed-loop, Conflict Resolution mode, fivesectors: a) aircraft count, b) minimum separation metric.


to a climbing aircraft as an added safety measure only to cancel it andallow the aircraft to continue climbing uninterrupted through thetemporary altitude.

The flight plan amendment metric for the closed-loop ConflictResolution mode simulation is shown in Figure 8. The same filteringwas applied to the closed-loop sample as was applied to the livesample to achieve an objective comparison of today’s operation versussimulation operations. Since under Control by Exception operationsaircraft were not deviated from their nominal route or altitude profilefor any reason other than to resolve a conflict, amendment activityshown in Figure 8 can be viewed as a measure of the minimumnumber of amendments needed to maintain safe separation. Theamendment activity in Figure 8 shows some correlation with theopen-loop separation characteristics shown in Figure 5. There wereno more than three amendments issued during any five min timeinterval, nor were there more than three aircraft with a minimumseparation of less than 5 nmi. The substantial decrease in amend-ments under Control by Exception (Figure 8) compared with today’soperations (Figure 7) is primarily due to the relatively large numberof temporary altitude amendments issued under today’s operations,and the fact that there were no direct route amendments issuedduring Conflict Resolution mode simulations. Under today’s opera-tions direct route amendments are relatively common in this airspaceand are addressed further in the next section.

Figure 7. Flight plan amendment metric, five sectors, live traffic baseline.

MCNALLY AND GONG52

Conflict Resolution and Direct-To Mode—Today’s FortWorth Center Traffic

In today’s operations, standardized routings are used to separatedeparture and arrival routes serving busy airports. This helps ensurethat controller workload does not exceed safe levels during heavyarrival and/or departure flows. These procedures often result in dog-legged routings that may not always be necessary, depending ontraffic conditions. In this section, the CTAS Direct-To function [Erz-berger et al., 1999] is used in the simulation to analyze the separa-tion characteristics and trajectory efficiency improvements thatcould be achieved if standard routings could be reduced or elimi-nated.

A closed-loop simulation was conducted using the five-sector ZFWairspace and the 90 min traffic sample described in the previoussection. The simulation was run in Conflict Resolution & Direct-Tomode where one controller (a NASA engineer) uses the trial plannerto resolve all conflicts and issue all conflict-free Direct-To routeamendments as soon as they appear on the Direct-To route advisorylist. The 24 s wilco time delay was applied. The resulting traffic countand minimum separation metric are shown in Figure 9. Figure 9bclearly shows that the separation characteristics for this run arecomparable to that of the baseline operations (Figure 4) and theConflict Resolution simulation (Figure 6). During this run, Direct-To

Figure 8. Flight plan amendment metric, five sectors, closed-loop, Conflict Resolu-tion mode.


route amendments were issued to 43 of the 167 simulated aircraft(26%) that flew through the airspace.

The net improvement in route efficiency was determined by apply-ing the route efficiency metric to the baseline and closed-loop trafficdata for this run. The route efficiency metric, as described earlier,measures the aggregate difference in path distance and flight timefor a common traffic sample flying in a common airspace but underdifferent operational procedures. In this case, the difference betweentoday’s operations (baseline) and the closed-loop Conflict Resolution& Direct-To mode simulation was measured. The difference in pathdistance and flight time for each aircraft was computed and accumu-lated as each aircraft exited the simulation airspace. For reference,the cumulative savings were plotted relative to the closed-loop exittime. Figure 9c shows the cumulative results of the route efficiency

Figure 9. Separation characteristics, closed-loop, Conflict Resolution & Direct-Tomode, five sectors: a) aircraft count, b) minimum separation metric, c) route effi-ciency metric.

MCNALLY AND GONG54

analysis of the 76 aircraft that met the initial and exit point condi-tions defined for the route efficiency metric. As the simulation pro-gressed the path distance and flight time savings increased as thecontroller issued Direct-To routes. As shown in Figure 9c, the cumu-lative flight time savings for all aircraft was 28 min. This equates to1.9 percent of the flight time within the simulation region, e.g., theflight time from IP to Exit-A in Figure 3. For those aircraft thatreceived Direct-To amendments, the flight time savings was 5.2 per-cent of the flight time within the simulation region. The results showthat an efficiency improvement was achieved while maintainingseparation assurance characteristics consistent with both today’s op-erations (Figure 4b) and automated operations without improvedrouting (Figure 6b). The improved route efficiency was attributed tothe TBA’s ability to identify and safely circumvent the proceduraldelays inherent to today’s sector-based operations.

The flight plan amendment metric for the closed-loop simulationconducted in the Conflict Resolution & Direct-To mode is shown inFigure 10. In this mode, the controller resolved all predicted conflictsand issued all Direct-To routes as described above. The direct routeamendments issued in this scenario may be considered a rough ap-proximation of pilot-requested plus controller-initiated direct routeamendments under today’s operations. Amendment activity in Fig-ure 10 reflects direct route amendments plus amendments issuedto maintain separation. The single controller operating with TBA

Figure 10. Flight plan amendment metric, five sectors, closed-loop, Conflict Reso-lution & Direct-To mode.


and data link communications issued an average of 4 amendmentsevery 5 minutes, three fewer than the live traffic baseline shown inFigure 7. This difference may actually be larger since vectors or al-titude changes that may have been issued verbally, but not enteredinto the Host computer, would not have been counted in the livetraffic baseline. There were significantly fewer altitude amendmentsissued during the simulation, 17 compared to 84 for the live trafficbaseline. This difference may be attributed to the increased precisionand efficiency offered by trajectory-based conflict detection algo-rithms, minimizing the need to issue temporary altitudes as addi-tional safety precautions as described above. There was approxi-mately the same number of route amendments issued during simu-lation, indicating the single simulation controller was able toaccommodate a comparable number of pilot requests handled by thelive traffic controllers and still maintain safe separation between allaircraft.

Conflict Resolution Mode—Two-Times Today’s Fort WorthCenter Traffic

In this section the traffic load was nearly doubled to evaluate theability of the TBA to enable improved efficiency and increased air-space capacity. The same five-sector ZFW airspace used in the pre-vious simulations of today’s traffic was used. The increased trafficlevel was achieved by combining recordings of two different trafficsamples from the same airspace, but at different times of day. Amorning traffic recording for a 90 min period starting at 0830 CST onJune 2, 2005 was combined with the late afternoon baseline record-ing used in the previous simulations. A filtering process was used toensure all aircraft were legally separated for their first two min inthe simulation airspace. If an aircraft pair was not initially legallyseparated, one aircraft of the pair was deleted from the scenario. Thisfiltering is consistent with today’s operations, since transfer of sepa-ration assurance responsibility for an aircraft from an upstream con-troller to a downstream controller is typically not initiated when theupstream controller is aware of an impending conflict involving theaircraft in the downstream controller’s airspace. In addition, anyduplicate aircraft call signs were modified to avoid confusion. Thismethod of increasing traffic load proved to be simple and effective forthe purposes of this study. However, it was not based on rigorousprojections of future traffic load and routing.

The minimum separation metric for the open-loop two-times nomi-nal traffic simulation is shown in Figure 11. As expected, the numberof aircraft pairs that pass with a minimum separation of less than thelegal minimum of 5 nmi was substantially higher than in the open-loop nominal traffic simulation shown in Figure 5.

MCNALLY AND GONG56

A closed-loop simulation was performed using the Conflict Resolu-tion mode. The traffic count and minimum separation metric for therun are shown in Figure 12. Figure 12b shows that under two-timesnominal traffic there were usually about 1 or 2 independent aircraftpairs in the five-sector airspace region that were flying near legalseparation criteria, i.e., with between 5 and 10 nmi separation.Clearly there were important exceptions to this norm, e.g., 7 aircraftpairs with 5–10 nmi minimum separation in the 75–80 min elapsedtime period and one instance where legal separation was lost in the30–35 min elapsed time period.

Any loss of separation is unacceptable in air traffic operations, sothe cause of the loss of separation in Figure 12b was investigated.Post simulation analysis of the encounter revealed that, due to anerror in the climb trajectory prediction for one of the aircraft, theconflict was not detected with enough lead time to resolve the conflictand prevent the loss of separation. Figure 13 shows that the nominalstrategic climb trajectory under-predicted the actual climb rate of oneof the aircraft (AAL708). Due to uncertainties in aircraft weight andclimb speed, the climb trajectory predictions are the most challengingfor today’s 4D trajectory analysis methods.

The encounter was re-played using the tactical detection function[Paielli and Erzberger, 2005], which simultaneously probes both thestrategic trajectory and the tactical trajectory and it was found that

Figure 11. Separation characteristics, two-times nominal traffic, open-loop, fivesectors: a) aircraft count, b) minimum separation metric.


the conflict was detected at 3 min before loss of separation. Three minwould have allowed adequate time for the controller to have resolvedthe conflict. Figure 13 shows that the tactical trajectory for AAL708better estimates the actual climb rate at this instance, resulting in anearlier conflict detection. (At the time of these simulations the tacti-cal detection algorithm was not running simultaneously with thestrategic conflict detection algorithms.) This loss of separation ex-ample and the post simulation analysis supports the concept of si-multaneous analysis of strategic and tactical trajectories for separa-tion assurance in trajectory-based operations with higher levels ofautomation [Erzberger, 2004].

CONCLUDING REMARKS

The conflict detection and trial planner resolution functions in theCenter/TRACON Automation System were configured to examinehow four-dimensional trajectory analysis methods could be extendedto support higher levels of automation for separation assurance inthe National Airspace System.

Human-in-the-loop laboratory simulations, typically 90 min in du-ration, were conducted where a human controller (a NASA engineer)manually generated conflict resolution trajectories using the inter-

Figure 12. Separation characteristics, two-times nominal traffic, closed-loop, Con-flict Resolution mode, five sectors: a) aircraft count, b) minimum separation metric.

MCNALLY AND GONG58

active trial planner resolution function, but only in response to con-flicts detected and displayed by the conflict detection function. Reso-lution trajectories were issued to simulated aircraft via a simulateddata link. Simulated aircraft automatically responded to resolutiontrajectories following a fixed 24 s compliance delay. Simulations werebased on actual FAA traffic data from the Fort Worth Center. Allconflict resolution trajectories were generated without considerationof routing restrictions or sector boundary considerations common intoday’s operations.

A single controller maintained legal separation (5 nmi horizontalor 1000 ft vertical) and improved the flying time efficiency by 1.9%while working the combined traffic in five Fort Worth Center high-altitude sectors at traffic levels nearly equivalent to that of today’straffic. Under laboratory conditions, the controller was performingthe separation assurance functions that are performed by 4–10people under today’s operations.

During a five-sector simulation at today’s traffic levels all aircraftthat could save at least 1 min or more of wind-corrected flight time byflying conflict-free direct routes to a downstream fix on their route offlight were immediately given direct route amendments without re-gard for today’s standard departure routings or inter-sector coordi-nation considerations. The controller maintained legal separationand issued conflict-free direct route amendments while working thecombined traffic in five Fort Worth Center high-altitude sectors attraffic levels nearly equivalent to that of today’s traffic. The improve-

Figure 13. Tactical trajectory analysis predicts loss of separation case in two-timesnominal run.


ment in flying time efficiency for the direct route aircraft alone was5.2%.

The results suggest that the use of trajectory-based automationhas the potential to substantially reduce the number of altitudeamendments required to ensure separation under today’s traffic lev-els.

During a simulation run at nearly two-times today’s traffic levels,in the combined five-sector Fort Worth Center airspace, a single con-troller maintained separation in all but one instance. A post-simulation analysis showed that a tactical alerting function pre-dicted the loss of separation at 3 min prior to loss of separation. Threeminutes is generally considered enough time for a controller to re-solve a conflict and prevent a loss of separation.

A trajectory uncertainty analysis showed that the trajectory pre-diction uncertainty associated with the simulations in this study,where traffic flows were initialized with actual FAA traffic data, isroughly consistent with the trajectory prediction uncertainty associ-ated with more realistic conditions, where predictions are based fullyon FAA traffic data. This makes the results more meaningful, sinceuncertainty is an unavoidable aspect of trajectory-based operations.

APPENDIX—TRAJECTORY PREDICTION UNCERTAINTY

A quantitative comparison of the trajectory prediction uncertaintyassociated with a given actual Center traffic sample with that of asimulated traffic sample initialized with the actual Center trafficsample provides insight as to the applicability of simulation resultsto the real-world environment. The trajectory accuracy analysismethodology described in [Gong and McNally, 2004] uses large num-bers of trajectory predictions to provide an aggregate measure ofaccuracy, which reflects the combined effect of all errors sources suchas wind, speed, and weight. Error characteristics are categorized asa function of flight phase (level, climbing, descending) and look-ahead time. The methodology was applied to the actual traffic sampleand to the open-loop simulated traffic sample for one of the 90 min5-sector Fort Worth Center traffic samples used in this analysis.The histogram in Figure 14 shows the level-flight trajectory errorcharacteristics at a 10 min look-ahead time for the actual trafficsample. Based on analysis of 185 level flight actual traffic trajectorypredictions, the mean along-track prediction error is −0.7 nmiand the standard deviation is 3.2 nmi. Figure 15 shows the corre-sponding along-track error characteristics for the simulated trafficsample that was only initialized with the actual traffic sample (173level flight trajectory predictions, mean error � 1.4 nmi, standard

MCNALLY AND GONG60

deviation � 2.9 nmi). Though the actual traffic errors appear a littlemore normally distributed than the simulated traffic errors, the over-all error characteristics are roughly similar. The climb trajectoryerror characteristics were analyzed and presented in the originalversion of this paper [McNally and Gong, 2006].

ACRONYMS

CDTI Cockpit Display of Traffic InformationCST Central Standard TimeCTAS Center/TRACON Automation SystemDFW Dallas/Fort Worth International AirportDST Decision Support ToolFAA Federal Aviation AdministrationIP Initialization PointNAS National Airspace SystemNASA National Aeronautics and Space AdministrationNGATS Next Generation Air Transport SystemNOAA National Oceanic and Atmospheric AdministrationRUC Rapid Update Cycle

Figure 14. Level flight trajectory accuracy, actual traffic baseline, along track er-ror, 10 min look-ahead.

Figure 15. Level flight trajectory accuracy, simulated traffic, along track error, 10min look-ahead.


TBA Trajectory-Based AutomationTRACON Terminal Radar Approach ControlZFW Fort Worth Center4D Four-Dimensional

REFERENCES

Barmore, B., Abbott, T., Krishnamurthy, K. (2004), “Airborne-Managed Spacing inMultiple Arrival Streams,” 24th International Congress of the Aeronautical Sci-ences (ICAS 2004), Yokohama, Japan, August 29–September 3, 2004.

Battiste, V., Johnson, W.W., Holland, S. (2000), “Enabling Strategic Flight DeckRoute Re-Planning Within A Modified ATC Environment: The Display of 4-DIntent Information on a CSD,” Proceedings of the 2000 SAE World Aviation Con-gress and Exposition, San Diego, CA.

Coppenbarger, R.A. (1999), “Climb Trajectory Prediction Enhancement Using Air-line Flight-Planning Information” AIAA-99–4147, AIAA Guidance, Navigation,and Control Conference, Portland, Oregon, August 1999.

Dekker, S.W.A., Woods, D.D. (1999), “To Intervene or not to Intervene: The Dilemmaof Management by Exception,” Cognition, Technology & Work (1999) 1:86–96,Springer-Verlag London Limited.

Erzberger, H. (2004), “Transforming the NAS: The Next Generation Air Traffic Con-trol System,” 24th International Congress of the Aeronautical Sciences (ICAS2004), Yokohama, Japan, August 29–September 3, 2004.

Erzberger, H., Davis, T.J., Green, S.M. (1993), “Design of Center-TRACON Automa-tion System,” Proceedings of the AGARD Guidance and Control Panel 56th Sym-posium on Machine Intelligence in Air-Traffic Management, Berlin, GDR, 1993.

Erzberger, H., McNally, D., Foster, M., Chiu, D., Stassart, P. (1999), “Direct-To Toolfor En Route Controllers,” ATM99: IEEE Workshop on Advanced Technologiesand their Impact on Air Traffic Management in the 21st Century, Capri, Italy, Sept26–30, 1999.

Erzberger, H., Paielli, R.A., Isaacson, D.R., Eshow, M.E. (1997), “Conflict Detectionand Resolution in the Presence of Prediction Error,” 1st USA/Europe Air TrafficManagement R&D Seminar, Saclay, France, June 17–20, 1997.

Federal Aviation Administration (2003), “Free Flight Program Performance MetricsResults to Date,” December 2003 Report.

Gong, C., McNally, D. (2004), “A Methodology for Automated Trajectory PredictionAnalysis,” AIAA 2004–4788, AIAA Guidance, Navigation, and Control Conference,Providence, RI, August 16–19, 2004.

Isaacson, D.R., Erzberger, H. (1997), “Design of a Conflict Detection Algorithm forthe Center/TRACON Automation System,” 16th Digital Avionics Systems Confer-ence, Irvine CA, October 26–30, 1997.

Joint Planning and Development Office (2004), “Next Generation Air TransportationSystem Integrated Plan,” December 12, 2004.

McNally, D., Bach, R., Chan, W. (1998), “Field Test Evaluation of the CTAS ConflictPrediction and Trial Planning Capability,” AIAA 99–4480, AIAA Guidance, Navi-gation and Control Conference, Boston MA, August 10–12, 1998.

McNally, D., Erzberger, H., Bach, R. (1999), “A Controller Tool for Transition Air-space,” AIAA 99–4298, AIAA Guidance, Navigation and Control Conference, Port-land, OR, August 9–11, 1999.

McNally, B.D., Engelland, S., Bach, R., Chan, W., Brasil, C., Gong, C., Frey, J.,Vincent, D. (2001), “Operational Evaluation of the Direct-To Controller Tool,” 4th

MCNALLY AND GONG62

USA/Europe Air Traffic Management R&D Seminar, Santa Fe, New Mexico, USA,December 3–7, 2001.

McNally, D., Gong, C. (2006), “Concept and Laboratory Analysis of Trajectory-BasedAutomation for Separation Assurance,” AIAA 2006–6600, AIAA Guidance, Navi-gation, and Control Conference, Keystone, CO, August 21–24, 2006.

Paielli, R.A., Erzberger, H. (1997), “Conflict Probability Estimation for Free Flight,”J. Guidance, Control, and Dynamics, Vol. 20, no 3, pp. 588–596, May-June, 1997.

Paielli, R.A., Erzberger, H. (2005), “Tactical Conflict Detection Methods for ReducingOperational Errors,” Air Traffic Control Quarterly, Vol. 13 (1) (2005).

Prevot, T., Lee, P., Smith, N., Palmer, E. (2005), “ATC Technologies for Controller-Managed and Autonomous Flight Operations,” AIAA Guidance, Navigation, andControl Conference, San Francisco, CA, 15–18 August 2005.

Robinson, J.E., Isaacson, D.R. (2002), “Development of a Closed-loop Testing Methodfor a Next-Generation Terminal Area Automation System,” American ControlConference, Anchorage, AK, May 8–10, 2002.

Schwartz, B.E., Benjamin, S.G., Green, S.M., Jardin, M.R. (2000), “Accuracy ofRUC-1 and RUC-2 Wind and Aircraft Trajectory Forecasts by Comparison withACARS Observations,” Weather Forecasting, 15, 313–326, June 2000.

Slattery, R.A., Zhao, Y. (1995), “En-Route Trajectory Synthesis for Air Traffic Con-trol Automation,” Proceedings of the 1995 American Control Conference, SeattleWA, June 21–23, 1995.

Swenson, H.N., Hoang, T., Engelland, S., Vincent, D., Sanders, T., Sanford, B.,Heere, K. (1997), “Design and Operational Evaluation of the Traffic ManagementAdvisor at the Fort Worth Air Route Traffic Control Center,” 1st USA/Europe AirTraffic Management R&D Seminar, Saclay, France, June 1997.

Weske, R.A., Danek, G.L. (1993), “Pseudo Aircraft Systems: A Multi-Aircraft Simu-lation System for Air Traffic Control Research,” Proceedings of the AIAA FlightSimulation Technologies Conference, Monterey, CA, August 1993.

Wing, D.J., Ballin, M.G., Krishnamurthy, K. (2004), “Pilot in Command: A Feasibil-ity Assessment of Autonomous Flight Management Operations,” 24th Interna-tional Congress of the Aeronautical Sciences (ICAS 2004), Yokohama, Japan, Au-gust 29–September 3, 2004.

BIOGRAPHIES

David McNally is a Principal Investigator for air traffic management research atNASA Ames Research Center. He has worked over the past twelve years in labora-tory development and analysis, human-in-the-loop simulation evaluation and opera-tional testing of conflict detection and resolution automation for air traffic manage-ment. Previously he conducted flight research on the use of satellite navigation forprecision landing guidance. He is currently the Associate Principal Investigator forSeparation Assurance under NASA’s research in support of the Next Generation AirTransportation System. He holds an MS degree in Mechanical Engineering and a BSdegree in Mechanical/Aeronautical Engineering, both from the University of Cali-fornia, Davis.

Chester Gong has been conducting air traffic management research at NASA AmesResearch Center for the past six years. He has previously worked as an engineer forUnited Airlines and the Air Force Flight Test Center. He holds a BS degree inAerospace Engineering from the California State Polytechnic University, Pomona.


concept and laboratory analysis of trajectory-based ...€¦ · analysis of trajectory-based...

Documents